Accuracy of Segmental Multi-Frequency Bioelectrical

ORIGINAL ARTICLE Accuracy of segmental multi-frequency bioelectrical impedance analysis for assessing whole-body and appendicular fat mass and lean soft tissue mass in frail women aged 75 years and older

M Kim and H Kim

BACKGROUND/OBJECTIVE: We aimed to examine the accuracy of segmental multi-frequency bioelectrical impedance analysis (SMF-BIA) for the assessment of whole-body and appendicular fat mass (FM) and lean soft tissue mass (LM) in frail older women, using dual-energy X-ray absorptiometry (DXA) as a reference method. SUBJECTS/METHODS: All 129 community-dwelling Japanese frail older women with a mean age of 80.9 years (range, 75–89 years) from the Frailty Intervention Trial were recruited. The agreements between SMF-BIA and DXA for whole-body and appendicular body composition were assessed using simple linear regression and Bland–Altman analysis. RESULTS: High coefficients of determination (R2) for whole-body FM (R2 ¼ 0.94, s.e. of estimate (SEE) ¼ 1.2 kg), whole-body LM (R2 ¼ 0.85, SEE ¼ 1.4 kg), and appendicular FM (R2 ¼ 0.82, SEE ¼ 1.1 kg) were observed between SMF-BIA and DXA. The R2 coefficient for appendicular LM was moderate (R2 ¼ 0.76, SEE ¼ 0.8 kg). Bland–Altman plots demonstrated that there was systematic (constant) bias (that is, DXA minus SMF-BIA) with overestimation of whole-body FM (bias ¼À1.2 kg, 95% confidence interval (CI) ¼À1.5 to À 0.1) and underestimation of whole-body LM (bias ¼ 2.1 kg, 95% CI ¼ 1.8–2.3) by SMF-BIA. Similar, the appendicular measurements also demonstrated systematic bias with overestimation of appendicular FM (bias ¼À0.3 kg, 95% CI ¼À0.5 to À 0.1) and underestimation of whole-body LM (bias ¼ 1.5 kg, 95% CI ¼ 1.4–1.7) by SMF-BIA. In addition, the individual level accuracy demonstrated a non-proportional bias for whole-body LM (r ¼ 0.08, P ¼ 0.338) and appendicular FM (r ¼ 0.07, P ¼ 0.413). CONCLUSIONS: SMF-BIA had acceptable accuracy for the estimation of whole-body and appendicular FM and LM in frail older women, although SMF-BIA underestimated LM and overestimated FM relative to DXA.

European Journal of Clinical Nutrition (2013) 67, 395–400; doi:10.1038/ejcn.2013.9; published online 6 February 2013 Keywords: body composition; bioelectrical impedance analysis; sarcopenia; frailty

INTRODUCTION and strength are a major component in the development of Frailty is an important and common geriatric syndrome that is frailty in the elderly.8,9 Moreover, frailty is associated with a described as a status of increased vulnerability resulting from the decline in muscle mass and quality and a parallel increase in fat loss of complexity in resting dynamics involving multiple mass (FM).10 Measurement of body composition, including FM physiological systems with advancing age.1 The prevalence of and muscle mass in older populations provide important frailty increases with age, from 3.9% at 65–74 years to 11.6% at information about their nutritional status. Therefore, the 75–84 years and to 25% in people older than 85 years. In addition, understanding of the body composition of frail elderly frailty is more prevalent in women than in men.1 Sarcopenia is a populations is an important part of clinical assessment with a loss of skeletal muscle mass and size that occurs with aging.2 goal of optimal prevention and treatment strategies. Although many definitions of sarcopenia have been reported,3–5 Dual-energy X-ray absorptiometry (DXA) is an accepted method current definitions focus on loss of appendicular skeletal muscle for the estimation of whole-body and segmental body fat and fat- mass as well as low muscle strength and low physical free mass (FFM), which includes lean soft tissues and bone performance.6 The European Working Group on Sarcopenia in minerals.11–13 However, DXA has disadvantages for use in clinical Older People consensus definition of sarcopenia is based on three settings, such as the high cost of equipment, risk of radiation stages: the presarcopenia stage involves low muscle mass with exposure and lack of access to instruments. For clinical use, normal muscle strength and physical performance; the bioelectrical impedance analysis (BIA) has been used as an sarcopenia stage involves low muscle mass and either attractive alternative method.4,14,15 BIA is a portable, non- diminished muscle strength or physical performance; and invasive, easy to use and convenient method for the patient, severe sarcopenia combines all three factors.6 Several and it is also relatively inexpensive compared with other pathophysiological overlaps between sarcopenia and frailty methods.16 Of the BIA devices developed over the years, have been observed.7 Thus, age-related loss in muscle mass segmental multi-frequency (SMF)-BIA devices have advantages

Research Team for Promoting Independence of the Elderly, Tokyo Metropolitan Institute of Gerontology, Tokyo, Japan. Correspondence: Dr M Kim, Research Team for Promoting Independence of the Elderly, Tokyo Metropolitan Institute of Gerontology, 35-2 Sakae-cho, Itabashi-ku, Tokyo 173-0015, Japan. E-mail: [email protected] Author contributions: Both authors designed the study together. MK developed the study concept and design, analysed and interpreted the data, and prepared the manuscript. HK recruited subjects, assisted with statistical analysis and reviewed the manuscript for accuracy. Received 17 September 2012; revised 19 December 2012; accepted 21 December 2012; published online 6 February 2013 Body composition estimation in the frail older women M Kim and H Kim 396 over single-frequency BIA devices (50 kHz).17–19 SMF-BIA avoids mineral density. The subjects were positioned for whole-body scans the problems encountered in single-frequency BIA by employing according to the manufacturer’s protocol. The subjects lay in a supine both low- and high-frequency electric currents.20 In recent years, position on the scanner table with their limbs close to their bodies. Their SMF-BIA has been shown to be valid in the estimation of body body compositions were analysed manually using DXA analysis software composition using DXA as a reference standard.21–23 However, (version 9.03 D; Hologic, Waltham, MA, USA). The segmental analyses of the total body into arm, leg and trunk segments were separated manually these results were obtained from analysis of healthy populations. with anatomical landmarks by the DXA analysis software. Appendicular To our knowledge, SMF-BIA has not been evaluated in the skeletal muscle mass25 was calculated as the sum of the LM of both the assessment of total and appendicular body composition in a right and left arms and legs, with the assumption that all non-fat and non- specifically targeted frail elderly population. Therefore, the aim of bone tissue was skeletal muscle. Appendicular muscle index was defined this study was to examine the accuracy of SMF-BIA for the as ASM/body height.23 The subjects were measured while wearing only a assessment of whole-body and appendicular body composition standard light cotton shirt to minimise clothing absorption. The DXA using DXA as a reference method in frail Japanese women aged machine was calibrated daily against a spine phantom supplied by the 75 years and older. manufacturer before testing. In addition, weekly calibration procedures were performed on a density step phantom. The precision error for bone mineral density and bone mineral content were 0.20–0.77% for the spine phantom. Our laboratory assessment of seven subjects demonstrated that MATERIALS AND METHODS the coefficients of variation for FM, LM and bone mineral content with Subjects repeated examinations were o3.0%. The subjects were 129 community-dwelling Japanese frail older women with a mean age of 80.9 years (range, 75–89 years). The study population was recruited from participants in the Frailty Intervention Trial (clinical Segmental multi-frequency bioelectrical impedance analysis trials registry, number: JMA-IIA00069). The baseline assessment was (SMF-BIA) conducted on 1835 women aged 75 and older at the Tokyo Metropolitan SMF-BIA was performed with a body composition analyser (InBody 720, Institute of Gerontology. Three hundred thirty-one were defined as frail, Biospace Co. Ltd, Soul, Korea). A tetra-polar 8-point tactile electrode system according to Fried’s frailty phenotype with the presence of three or more was used. The system separately measured the impedance of the subjects’ of following criteria: weight loss, weakness, exhaustion, slowness and low right arm, left arm, trunk, right leg and left leg at six different frequencies physical activity.1 In the present study, the five different components of the (1, 5, 50, 250, 500 and 1000 kHz) for each body segment. In accordance frailty indicators were evaluated as: (1) weight loss: either, answering ‘yes’ with the manufacturer’s guidelines, subjects wiped the bottom of their feet to the question, ‘In the last 6 months, have you lost 42B3kg with a proprietary electrolyte tissue before standing on the electrodes unintentionally?’ or a body mass index (BMI) o18.5; (2) weakness: hand embedded in the scale platform of the respective analysers. The subjects grip strength o19.0 kg; (3) slowness: usual walking speed o1.10 m/s; were instructed to stand upright and to grasp the handles of the analyser, (4) exhaustion: answering ‘yes’ to at least one of two questions, ‘I felt that thereby providing contact with a total of eight electrodes (two for each everything I did was an effort’ or ‘I could not get going’; (5) low physical foot and hand). In our study, the within-day coefficient variances for six activity: answering ‘true’ to at least three of the following four statements, different frequencies evaluated in nine subjects were 0–1.9%. Proprietary ‘I regularly take walks less than once a week,’ ‘I do not exercise regularly,’ equations from the manufacturer were used to estimate whole and ‘I do not actively participate in hobbies or lessons of any sort,’ and ‘I do not regional body composition variables. participate in any social groups for elderly people or volunteering.’ Two hundred (60.4%) of the frail older women were excluded because they were classified into the exclusion criteria or declined participation. The Statistical analysis exclusion criteria were: (1) severe knee or back pain; (2) severely impaired The data are expressed as means, s.d. , and range (minimum-maximum). A mobility; (3) impaired cognition (Mini-Mental State Examination score paired Student’s t-test was used to compare the difference in body o24); (4) missing baseline data; and (5) unstable cardiac conditions, such composition measurements between the SMF-BIA and DXA. To assess the as ventricular dysrhythmias, pulmonary oedema or other musculoskeletal agreement in body composition parameters of whole-body measurements conditions. Of a total of 131 frail older women who participated in the of FM and LM and appendicular measurements of FM and LM as measured intervention study, body composition was measured in 129 subjects, using by SMF-BIA and DXA, linear regression and Bland–Altman analyses were SMF-BIA and DXA. The anthropometric assessment of the subjects was conducted. Simple linear regression analyses were performed with DXA conducted at the Tokyo Metropolitan Institute of Gerontology. The body composition parameters as the dependent variable to determine participants read and signed the informed consent forms that were whether the regression line differed significantly from the line of identity. 26 approved by the institutional review board before testing. The Clinical In the Bland–Altman plots, the systematic bias was calculated as the Research Ethics Committee of the Tokyo Metropolitan Institute of mean difference between methods, and the 95% limits of agreement were Gerontology approved the study protocol. calculated as the bias ±2 s.d. of the differences between methods. As there was evidence of proportional bias for body composition parameters, a Pearson’s correlation was performed to quantify the bias observed in the Experimental design Bland–Altman plots. Multiple regression analysis was performed to The study model was a cross-sectional analysis of baseline data from the determine physical variables that influenced the bias of appendicular LM Frailty Intervention Trial. The subjects were instructed to refrain from between DXA and SMF-BIA. The independent variables were age, body exercise for 12 h and to refrain from eating for 3 h and drinking for 1 h weight, height and appendicular LM as determined by DXA. Statistical bfore the measurements.24 Subject body composition was measured by analyses were performed using the IBM SPSS software version 20 SMF-BIA and DXA. Both investigations were performed on the same day (SPSS Inc., Chicago, IL, USA) and the SigmaPlot software version 12.0 2 h apart. (Systat Software Inc., Chicago, IL, USA). For all tests, statistical significance was set at Po0.05. Anthropometric measurements With the subjects wearing light clothes and no shoes, body weight was measured to the nearest 0.01 kg using DXA equipment, and height was RESULTS determined to the nearest 0.1 cm using a fixed-wall-scale measuring The characteristics of the frail older women subjects are described device. BMI was calculated as body weight in kilograms divided by height in Table 1 with means±s.d. and ranges. Table 2 describes the in metres squared. The calf circumference was measured at the point of body composition parameters obtained by using SMF-BIA and greatest circumference. DXA. The means of the body composition parameters estimated by SMF-BIA and DXA were significantly different (Po0.01), except Dual-energy X-ray absorptiometry (DXA) for the segmental FM in both legs (P40.05). As a reference method, DXA (QDR-4500 A scanner; Hologic, Waltham, MA, USA) Figure 1 displays the results of simple linear regression analyses was used for the measurement of whole and regional body composition, for whole-body FM and LM, in addition to the appendicular FM including FM, lean soft tissue mass (LM), bone mineral content and bone and LM parameters as determined by SMF-BIA and DXA. The

European Journal of Clinical Nutrition (2013) 395 – 400 & 2013 Macmillan Publishers Limited Body composition estimation in the frail older women M Kim and H Kim 397 correlations between SMF-BIA and the body composition confidence interval (CI) ¼ 1.5 to À 0.1) and the underestimation parameters estimated by DXA for whole-body FM and LM of whole-body LM (bias ¼ 2.1 kg, 95% CI ¼ 1.8–2.3) by SMF-BIA. and appendicular FM were high (r 40.9, all Po0.001). High Proportional bias was noted for whole-body FM measurement, coefficients of determination (R2) for whole-body FM (R2 ¼ 0.94, with overestimation of the whole-body FM (SMF-BIA) increasing s.e. of estimate (SEE) ¼ 1.2 kg or 8%), whole-body LM (R2 ¼ 0.85, with increasing whole-body FM (r ¼À1.42, Po0.01). However, SEE ¼ 1.4 kg or 6%) and appendicular FM (R2 ¼ 0.82, SEE ¼ 1.1 kg the Bland–Altman plots indicated no significant proportional bias or 15%) between SMF-BIA and DXA were observed. The in whole-body LM measurement (r ¼ 0.08, P ¼ 0.338). Similarly, the R2 coefficient for appendicular LM was moderate (R2 ¼ 0.76, appendicular parameters were systematically biased, with the SEE ¼ 0.8 kg or 6%). overestimation of appendicular FM (bias ¼À0.3 kg, 95% CI ¼ In addition, agreements between the two methods were À 0.5 to À 0.1) and the underestimation of whole-body LM assessed using Bland–Altman plots at the individual level (bias ¼ 1.5 kg, 95% CI ¼ 1.4–1.7) by SMF-BIA. In contrast, the (Figure 2). There was a narrow limit of agreement on the Bland– Bland–Altman plots indicated no significant proportional bias in Altman plots for the whole-body FM and LM and the appendicular appendicular FM measurement (r ¼ 0.07, P ¼ 0.413). In addition, FM and LM measurements. Almost all individual plots were within proportional bias was noted for appendicular LM measurement, the 95% limit of agreement (mean difference±2 s.d.). There was with SMF-BIA tending to underestimate the appendicular LM in systematic (constant) bias (that is, DXA minus SMF-BIA) with the the lower range (r ¼À1.42, Po0.01). overestimation of whole-body FM (bias ¼À1.2 kg, 95% In a multiple regression analysis, age (b ¼ 0.051), body weight (b ¼À0.055), height (b ¼À0.091) and appendicular LM as determined by DXA (b ¼ 0.302) were significant contributors to Table 1. Characteristics of the subjects the appendicular LM bias between DXA and SMF-BIA (all, Po0.05) (data not shown). The R2 in the multiple regression model was Mean±s.d. Range 0.421, indicating that 42.1% of the variability in the appendicular Age, years 80.9±2.9 75.0–89.0 LM bias was explained by all variables (P ¼ 0.001). Body weight, kga 48.5±8.2 29.2–72.4 Height, cm 146.4±6.0 132.2–161.8 BMI, kg/m2 22.6±3.5 15.6–31.4 o18.5 32 (24.8) DISCUSSION 18.5–24.9 80 (62.0) To our knowledge, this is the first investigation to compare the X25.0 17 (13.2) assessment of whole-body and appendicular body composition ± Calf circumference, cm 32.4 3.0 25.7–41.3 from SMF-BIA to DXA device-based measurements in a commu- o31.0 46 (35.7) Whole-body bone mineral 1111.1±254.0 978.1–1880.1 nity-dwelling elderly population of frail women Japanese aged 75 content, g years and older. In particular, our study examined the accuracy of Whole-body bone mineral 0.75±0.10 0.59–1.37 SMF-BIA in the heterogeneous population. Our findings indicate density, g/cm2 that there was good agreement between the two methods for the estimation of whole-body and appendicular body composition in Abbreviation: BMI, body mass index. frail older women subjects, but SMF-BIA underestimated LM and Values are means±s.d., number (%). aWeight derived from whole-body mass measurement by dual X-ray overestimated FM relative to DXA. Moreover, the Bland–Altman absorptiometry. plots at the individual level demonstrated non-proportional bias for whole-body LM and appendicular FM.

Table 2. Body composition parameters as determined by DXA and SMF-BIA

Body composition parameters DXA SMF-BIA Differencea

Mean±s.d. Range Mean±s.d. Range Mean±s.d. P-valueb

Whole-body measurement FM, kg 14.7±5.1 4.4–30.3 16.0±5.7 4.2–33.6 À 1.2±1.5 0.001 LM, kg 32.7±3.6 24.1–42.0 30.6±3.5 23.0–41.5 2.1±1.4 0.001 Percentage of FM, % 29.6±5.9 13.2–41.8 32.0±7.0 12.6–49.7 À 2.5±2.8 0.001

Segmental body mass measurement Right arm FM, kg 1.0±0.4 0.3–2.6 1.7±0.5 0.4–3.0 À 0.2±0.2 0.001 Left arm FM, kg 1.0±0.4 0.3–2.5 1.2±0.5 0.4–3.1 À 0.2±0.2 0.001 Trunk FM, kg 6.7±2.7 1.6–15.5 7.6±3.1 0.9–17.1 À 0.8±1.0 0.001 Right leg FM, kg 2.6±2.0 0.6–5.0 2.9±0.8 0.9–4.7 0.1±0.5 0.177 Left leg FM, kg 2.6±0.9 0.6–4.9 2.6±0.8 0.9–4.7 0.0±0.5 0.816 Appendicular FM, kg 7.2±2.6 1.8–13.6 7.5±2.5 2.6–15.2 À 0.3±1.1 0.001 Right arm LM, kg 1.6±0.2 1.1–2.2 1.4±0.3 0.7–2.1 0.2±0.2 0.001 Left arm LM, kg 1.6±0.2 1.0–2.2 1.4±0.3 0.70–2.1 0.2±0.2 0.001 Trunk LM, kg 16.4±2.0 11.7–21.7 13.7±2.0 9.0–18.2 2.7±1.0 0.001 Right leg LM, kg 5.1±0.6 3.8–6.9 4.5±0.8 2.9–7.0 0.6±0.4 0.001 Left leg LM, kg 5.1±0.6 3.7–7.1 4.5±0.8 3.0–7.2 0.6±0.4 0.001 Appendicular LM, kg 13.4±1.6 10.0–18.0 11.9±2.0 7.7–18.3 1.6±0.9 0.001 Appendicular skeletal muscle index, kg/m2c 6.3±0.7 4.8–8.1 5.5±0.7 4.0–7.9 0.8±0.5 0.001 Abbreviations: DXA, dual X-ray absorptiometry; FM, fat mass; LM, lean soft tissue mass; SMF-BIA, segmental multi-frequency bioelectrical impedance analysis. Values are means±s.d. aMean difference between DXA and BIA (that is, DXA minus SMF-BIA), mean (s.d.) bP-values for paired t-test between DXA and SMF-BIA. cAppendicular lean soft tissue mass (kg)/height (m2).

& 2013 Macmillan Publishers Limited European Journal of Clinical Nutrition (2013) 395 – 400 Body composition estimation in the frail older women M Kim and H Kim 398 40 45

40 30

35 20 30 Whole-body

10 Y = 0.88X + 0.88 Y = 0.95X + 3.58 r = 0.97 (R2 = 0.94) 25 r = 0.92 (R2 = 0.85) P < 0.001 P < 0.001 lean soft tissue mass by DXA (kg) Whole-body fat mass by DXA (kg) SEE = 1.2kg (8%) SEE = 1.4kg (6%) 0 20 01020 30 40 20 2530 35 40 45 Whole-body fat mass by SMF-BIA (kg) Whole-body lean soft tissue mass by SMF-BIA (kg)

20 16

14 18

12 16 10 14 8 12 6 Appendicular Y = 0.93X + 0.22 10 Y = 0.72X + 4.96 4 r=0.90 (R2 = 0.82) r = 0.87 (R2 = 0.76) P < 0.001 2 P < 0.001 8

SEE = 1.1kg (15%) lean soft tissue mass by DXA (kg) SEE = 0.8kg (6%) Appendicular fat mass by DXA (kg) 0 6 0421666108 10 12 14 8182012 14 16 Appendicular fat mass by SMF-BIA (kg) Appendicular lean soft tissue mass by SMF-BIA (kg) Figure 1. Linear regression between SMF-BIA and DXA. (a) Whole-body fat mass, (b) whole-body lean soft tissue mass, (c) appendicular fat mass and (d) appendicular lean soft tissue mass. SEE, s.e. of estimate; solid lines, regression line; dotted lines, identity line.

4 r = -0.42 8 P < 0.01 r = 0.08 2 6 P = 0.338 0 4 -2 2 -4

(DXA - SMF-BIA) (kg) 0 (DXA - SMF-BIA) (kg) -6 Difference between methods Difference between methods -8 -2 0 5 10 15 20 25 30 35 25 30 35 40 45 Average of whole-body Average of whole-body fat mass of two methods (kg) lean soft tissue mass of two methods (kg)

4 5 r = -0.42 3 r = 0.07 4 P < 0.01 P = 0.413 2 3 1 2 0 1 -1 0 -2 -1 (DXA - SMF-BIA) (kg) -3 (DXA - SMF-BIA) (kg) -2 Difference between methods -4 Difference between methods -3 0 2468101214 16 8 101214161820 Average of appendicular Average of appendicular fat mass of two methods (kg) lean soft tissue mass of two methods (kg) Figure 2. Bland–Altman plots comparing: (a) whole-body fat mass, (b) whole-body lean soft tissue mass, (c) appendicular fat mass and (d) appendicular lean soft tissue mass by SMF-BIA and DXA. Solid lines, bias (mean difference); dotted lines, limits of agreement (mean difference±2 s.d.).

Previous studies have demonstrated that SMF-BIA provides a Biospace Co. Ltd) had a good agreement with DXA (the same valid estimation of body composition using DXA as a reference device as used in this study, the Hologic QDR-4500A) in the standard.21–23 Ling et al.22 reported that SMF-BIA (InBody 720, assessment of total body composition of 484 general middle-aged

European Journal of Clinical Nutrition (2013) 395 – 400 & 2013 Macmillan Publishers Limited Body composition estimation in the frail older women M Kim and H Kim 399 Dutch subjects. In that study, the coefficients of determination for from healthy populations, they may contribute to error in body whole-body FM (R2 ¼ 0.94) and whole-body FFM (R2 ¼ 0.95) in composition measurements in specific populations. linear regression equations with adjusted gender was significantly This study measured coefficients of determination for appendi- greater. Anderson et al.18 found that whole-body FM (R2 ¼ 0.95) cular FM (R2 ¼ 0.82) and appendicular LM (R2 ¼ 0.76) between and LM (R2 ¼ 0.88) measured with SMF-BIA in 25 women aged SMF-BIA and DXA. Our findings are supported by previous studies 18–45 years had a high correlation and small SEE when using DXA that indicate SMF-BIA has excellent agreement in the measure- (Lunar DPX-iQ2288) as a reference standard. Houtkouper et al.27 ment of the segmental LM as both the right and left arms when reported that an SEE of 2.0–2.5 kg in men and 1.5–1.8 kg in women using DXA as the reference method (interclass correlation is considered ideal in the FFM as calculated by the BIA equations. coefficient X0.83).22 Anderson et al.21 found that the Whole-body LM, as measured by DXA, is bone mineral-free measurement of appendicular LM by SMF-BIA devices (InBody LM (total FFM À total bone mineral content). Previous studies 720 and InBody520) was moderately to strongly associated have reported good correlations between DXA-derived (R2 ¼ 0.62–0.87) with DXA in men and women aged 18–49. In LM and skeletal muscle mass when MRI was used as the our study, the appendicular FM was in better agreement between criterion (r ¼ 0.94–0.97).11,28–30 Chen et al.11 reported that DXA- SMF-BIA and DXA than the appendicular LM. To our knowledge, derived LM was highly correlated with MRI-derived whole-body no comparative studies exist that evaluate the accuracy of skeletal muscle mass (r ¼ 0.94) in 101 older women aged 50–79 assessing the segmental body composition at the individual level years. Our study found that the Bland–Altman plots indicated no by SMF-BIA (InBody 720 device) in a population of elderly subjects. significant proportional bias in whole-body LM measurement. In the present study, despite the significant SMF-BIA over- Therefore, SMF-BIA may provide a valid method for assessing estimation of appendicular FM and the underestimation of whole-body body composition, particularly for the whole-body appendicular LM with DXA, the Bland–Altman plots indicated a skeletal muscle mass, assuming that the LM from DXA is skeletal non-proportional bias in appendicular FM measurement. However, muscle mass in the frail older women. we observed a proportional bias in appendicular LM, with SMF-BIA We found in our study that SMF-BIA underestimated whole- tending to underestimate appendicular LM in the lower range (see body LM and overestimated whole-body FM relative to DXA (see Figure 2). These results are in contrast to the results of previous Figure 1). In our study, a subanalysis of the FFM indicated that studies evaluating SMF-BIA in healthy adults. Anderson et al.21 SMF-BIA underestimated the whole-body FFM (bias, 1.2 kg, 95% CI, found a non-proportional bias for appendicular LM as measured 0.9–1.5) (data not shown). These results are consistent with a by two types of SMF-BIA devices in 25 women with a mean BMI of previous study. The method’s bias indicated that SMF-BIA under- 26.1 kg/m2 and aged 18–45. These different findings are probably estimated whole-body FFM and overestimated whole-body FM in the result of methodological differences, with the previous data women with a mean age of 61.2±6.4 years and a mean BMI of confined to small subject numbers dispersed over a wide age 26.1±4.4 kg/m2.22 However, Vo¨lgyi et al.31 demonstrated the range. In particular, the findings may be the result of a validity of SMF-BIA compared with DXA (GE Lunar Prodig) in 86 combination of physical factors such as different body sizes. Finnish women aged 37–81. These researchers observed that Bedogni et al.17 found that eight-polar SMF-BIA was precise and SMF-BIA overestimated FFM in normal and overweight groups by gave accurate estimates of TBW in healthy subjects with a BMI 3.2 and 2.9 kg, respectively. Discrepancies between studies are range from 18.5–29.9 kg/m2. In our study population, the prevalence most likely due to differences in the specificity of subject of underweight subjects (BMI values below 18.5 kg/m2)inthefrail populations (for example, age, gender, body shape, ethnic older women population was 24%, with a TBW-to-body weight ratio groups). In our study, SMF-BIA was used to analyse body of 44.8%. Thus, the Fried’s definition includes weight loss criteria.1 composition (InBody 720 device). The measurement of FFM with We found that in multiple regression analysis, the age, body weight, an InBody 720 device was estimated as TBW/0.73. In addition, FM height and appendicular LM determined by DXA were associated was calculated as the difference between total body weight and with the bias of appendicular LM between DXA and SMF-BIA among FFM. However, FFM hydration of 0.73 has been shown to be the frail older women subjects. Therefore, SMF-BIA may tend to remarkably stable in healthy individuals.32 The change of FFM underestimate appendicular LM in the lower range as underweight hydration has been controversial because of the presence of when using DXA as the reference method. systematic differences in regards to growth, aging, adiposity, Our study has some limitations. First, although DXA is a gender, body size and acute or catabolic illness.33 Heymsfield validated ‘gold standard’ reference method, it is still only an et al.34 suggest that FFM hydration increases slightly in old age, estimate of body composition. Therefore, validation against DXA is resulting in a slight, systematic decrease in FFM density. not the most accurate analysis possible.42–44 However, it is Physiological ageing is associated with several changes that may included as a reference method because of its wide availability affect water balance and expose older adults to the risk of and previous validation. Second, it is likely that the focus of our dehydration. These changes include a decline in renal function study on frail older women in communities may not be applicable and thirst perception and a reduction of TBW.35 Thus, SMF-BIA to populations in nursing homes, hospitals and other institutions. may lead to underestimation of FFM with DXA in the dehydrated Finally, the hydration status of the study subjects was not state. The extracellular water to intracellular water (ECW/ICW) ratio determined before the body composition assessment. is a parameter of cellular hydration state. The ECW/ICW ratio In conclusion, the present study confirmed that SMF-BIA had ranges from 0.80–1.20 in healthy adults.33 However, elderly acceptable accuracy in the estimation of whole-body and patients displayed chronic cellular dehydration associated with appendicular FM and LM in frail women subjects aged 75 years relative extracellular overhydration, which was not evidently and older, although SMF-BIA underestimated LM and over- related to ageing because healthy elderly volunteers and estimated FM relative to DXA. In addition, the individual level healthy adults had similar water space distributions.36 Notably, accuracy revealed non-proportional bias for whole-body LM and overhydration is a frequent consequence of organ failures such as appendicular FM measurement. This may suggest that SMF-BIA can kidney impairment, heart failure, chronic obstructive pulmonary be used in intrapersonal comparisons, with the understanding that disease and liver disease.37–41 Basrends et al.38 reported that SMF-BIA measurements will include errors. Our findings indicate chronic obstructive pulmonary disease patients with extreme FFM that SMF-BIA would be useful for community-based research in wasting are characterised by an increased ECW/ICW ratio despite measuring body composition in frail older women populations. the relative sparing of FM. Therefore, SMF-BIA is dependent on Future research efforts should examine the validity of the SMF-BIA proprietary regression equations to estimate conductor volume models in predicting body composition changes in frail elderly (for example, FFM). As these equations have been formulated populations with diverse body shapes and compositions.

& 2013 Macmillan Publishers Limited European Journal of Clinical Nutrition (2013) 395 – 400 Body composition estimation in the frail older women M Kim and H Kim 400 CONFLICT OF INTEREST total and regional body composition varies between men and women. Nutr Res The authors declare no conflict of interest. 2012; 32: 479–485. 22 Ling CH, de Craen AJ, Slagboom PE, Gunn DA, Stokkel MP, Westendorp RG et al. Accuracy of direct segmental multi-frequency bioimpedance analysis in the ACKNOWLEDGEMENTS assessment of total body and segmental body composition in middle-aged adult We are deeply grateful to the study participants and to the staff of the Tokyo population. Clin Nutr 2011; 30: 610–615. Metropolitan Institute of Gerontology for their cooperation. 23 Shafer KJ, Siders WA, Johnson LK, Lukaski HC. Validity of segmental multiple- frequency bioelectrical impedance analysis to estimate body composition of adults across a range of body mass indexes. Nutrition 2009; 25: 25–32. REFERENCES 24 Kyle UG, Bosaeus I, De Lorenzo AD, Deurenberg P, Elia M, Manuel Gomez J et al. 1 Fried LP, Tangen CM, Walston J, Newman AB, Hirsch C, Gottdiener J et al. Frailty in Bioelectrical impedance analysis-part II: utilization in clinical practice. Clin Nutr older adults: evidence for a phenotype. J Gerontol A Biol Sci Med Sci 2001; 56: 2004; 23: 1430–1453. M146–M156. 25 Takebe K, Narayan MG, Kukulka C, Basmajian JV. Slowing of nerve conduction 2 Rosenberg IH. Summary comments. Am J Clin Nutr 1989; 50: 1231–1233. velocity in hemiplegia: possible factors. Arch Phys Med Rehabil 1975; 56: 285–289. 3 Baumgartner RN, Koehler KM, Gallagher D, Romero L, Heymsfield SB, Ross RR et al. 26 Bland JM, Altman DG. Statistical methods for assessing agreement between two Epidemiology of sarcopenia among the elderly in New Mexico. Am J Epidemiol methods of clinical measurement. Lancet 1986; 1: 307–310. 1998; 147: 755–763. 27 Houtkooper LB, Lohman TG, Going SB, Howell WH. Why bioelectrical impedance 4 Janssen I, Heymsfield SB, Ross R. Low relative skeletal muscle mass (sarcopenia) in analysis should be used for estimating adiposity. Am J Clin Nutr 1996; 64: older persons is associated with functional impairment and physical disability. 436S–448S. J Am Geriatr Soc 2002; 50: 889–896. 28 Freda PU, Shen W, Reyes-Vidal CM, Geer EB, Arias-Mendoza F, Gallagher D et al. 5 Roubenoff R. Sarcopenia: effects on body composition and function. J Gerontol A Skeletal muscle mass in acromegaly assessed by magnetic resonance imaging Biol Sci Med Sci 2003; 58: 1012–1017. and dual-photon x-ray absorptiometry. J Clin Endocrinol Metab 2009; 94: 6 Cruz-Jentoft AJ, Baeyens JP, Bauer JM, Boirie Y, Cederholm T, Landi F et al. 2880–2886. Sarcopenia: European consensus on definition and diagnosis: Report of the Eur- 29 Kim J, Shen W, Gallagher D, Jones Jr. A, Wang Z, Wang J et al. Total-body skeletal opean Working Group on Sarcopenia in Older People. Age Ageing 2010; 39:412– muscle mass: estimation by dual-energy X-ray absorptiometry in children and 423. adolescents. Am J Clin Nutr 2006; 84: 1014–1020. 7 Walston J, Hadley EC, Ferrucci L, Guralnik JM, Newman AB, Studenski SA et al. 30 Shih R, Wang Z, Heo M, Wang W, Heymsfield SB. Lower limb skeletal muscle mass: Research agenda for frailty in older adults: toward a better understanding of development of dual-energy X-ray absorptiometry prediction model. J Appl physiology and etiology: summary from the American Geriatrics Society/National Physiol 2000; 89: 1380–1386. Institute on Aging Research Conference on Frailty in Older Adults. J Am Geriatr Soc 31 Volgyi E, Tylavsky FA, Lyytikainen A, Suominen H, Alen M, Cheng S. Assessing 2006; 54: 991–1001. body composition with DXA and bioimpedance: effects of obesity, physical 8 Rolland Y, Czerwinski S, Abellan Van Kan G, Morley JE, Cesari M, Onder G et al. activity, and age. Obesity 2008; 16: 700–705. Sarcopenia: its assessment, etiology, pathogenesis, consequences and future 32 Wang Z, Deurenberg P, Wang W, Pietrobelli A, Baumgartner RN, Heymsfield SB. perspectives. J Nutr Health Aging 2008; 12: 433–450. Hydration of fat-free body mass: review and critique of a classic body-composi- 9 Roubenoff R. Sarcopenia: a major modifiable cause of frailty in the elderly. J Nutr tion constant. Am J Clin Nutr 1999; 69: 833–841. Health Aging 2000; 4: 140–142. 33 Wang Z, Deurenberg P, Wang W, Pietrobelli A, Baumgartner RN, Heymsfield SB. 10 Cesari M, Leeuwenburgh C, Lauretani F, Onder G, Bandinelli S, Maraldi C et al. Hydration of fat-free body mass: new physiological modeling approach. Am J Frailty syndrome and skeletal muscle: results from the Invecchiare in Chianti Physiol 1999; 276: E995–E1003. study. Am J Clin Nutr 2006; 83: 1142–1148. 34 Heymsfield SB, Nunez C, Testolin C, Gallagher D. Anthropometry and methods of 11 Chen Z, Wang Z, Lohman T, Heymsfield SB, Outwater E, Nicholas JS et al. Dual- body composition measurement for research and field application in the elderly. energy X-ray absorptiometry is a valid tool for assessing skeletal muscle mass in Eur J Clin Nutr 2000; 54(Suppl 3): S26–S32. older women. JNutr2007; 137: 2775–2780. 35 Buffa R, Floris GU, Putzu PF, Marini E. Body composition variations in ageing. Coll 12 Prior BM, Cureton KJ, Modlesky CM, Evans EM, Sloniger MA, Saunders M et al. In Antropol 2011; 35: 259–265. vivo validation of whole body composition estimates from dual-energy X-ray 36 Ritz P. Investigators of the Source S, of the Human Nutrition Research C-A. Chronic absorptiometry. J Appl Physiol 1997; 83: 623–630. cellular dehydration in the aged patient. J Gerontol A Biol Sci Med Sci 2001; 56: 13 Salamone LM, Fuerst T, Visser M, Kern M, Lang T, Dockrell M et al. Measurement of M349–M352. fat mass using DEXA: a validation study in elderly adults. J Appl Physiol 2000; 89: 37 Wizemann V, Wabel P, Chamney P, Zaluska W, Moissl U, Rode C et al. The mor- 345–352. tality risk of overhydration in haemodialysis patients. Nephrol Dial Transplant 14 Chien MY, Huang TY, Wu YT. Prevalence of sarcopenia estimated using a bio- 2009; 24: 1574–1579. electrical impedance analysis prediction equation in community-dwelling elderly 38 Baarends EM, Schols AM, van Marken Lichtenbelt WD, Wouters EF. Analysis of people in Taiwan. J Am Geriatr Soc 2008; 56: 1710–1715. body water compartments in relation to tissue depletion in clinically stable 15 Genton L, Graf CE, Karsegard VL, Kyle UG, Pichard C. Low fat-free mass as a marker patients with chronic obstructive pulmonary disease. Am J Clin Nutr 1997; 65: of mortality in community-dwelling healthy elderly subjects. Age Ageing 2012; 42: 88–94. 33–39. 39 Ritz P, Berrut G. The importance of good hydration for day-to-day health. Nutr Rev 16 Chumlea WC, Guo SS. Bioelectrical impedance and body composition: present 2005; 63: S6–13. status and future directions. Nutr Rev 1994; 52: 123–131. 40 Ritz P, Vol S, Berrut G, Tack I, Arnaud MJ, Tichet J. Influence of gender and body 17 Bedogni G, Malavolti M, Severi S, Poli M, Mussi C, Fantuzzi AL et al. Accuracy of an composition on hydration and body water spaces. Clin Nutr 2008; 27: 740–746. eight-point tactile-electrode impedance method in the assessment of total body 41 Sergi G, Lupoli L, Volpato S, Bertani R, Coin A, Perissinotto E et al. Body fluid water. Eur J Clin Nutr 2002; 56: 1143–1148. distribution in elderly subjects with congestive heart failure. Ann Clin Lab Sci 2004; 18 Pietrobelli A, Morini P, Battistini N, Chiumello G, Nunez C, Heymsfield SB. Appendicular skeletal muscle mass: prediction from multiple frequency seg- 34: 416–422. mental bioimpedance analysis. Eur J Clin Nutr 1998; 52: 507–511. 42 Goran MI, Toth MJ, Poehlman ET. Assessment of research-based body 19 Pietrobelli A, Nunez C, Zingaretti G, Battistini N, Morini P, Wang ZM et al. composition techniques in healthy elderly men and women using the Assessment by bioimpedance of forearm cell mass: a new approach to calibration. 4-compartment model as a criterion method. Int J Obes Relat Metab Disord 1998; Eur J Clin Nutr 2002; 56: 723–728. 22: 135–142. 20 Gibson AL, Holmes JC, Desautels RL, Edmonds LB, Nuudi L. Ability of new octa- 43 Roubenoff R, Kehayias JJ, Dawson-Hughes B, Heymsfield SB. Use of dual-energy polar bioimpedance spectroscopy analyzers to predict 4-component-model per- x-ray absorptiometry in body-composition studies: not yet a ‘gold standard’. Am J centage body fat in Hispanic, black, and white adults. Am J Clin Nutr 2008; 87: Clin Nutr 1993; 58: 589–591. 332–338. 44 Snead DB, Birge SJ, Kohrt WM. Age-related differences in body composition by 21 Anderson LJ, Erceg DN, Schroeder ET. Utility of multifrequency bioelectrical hydrodensitometry and dual-energy X-ray absorptiometry. J Appl Physiol 1993; 74: impedance compared with dual-energy x-ray absorptiometry for assessment of 770–775.

European Journal of Clinical Nutrition (2013) 395 – 400 & 2013 Macmillan Publishers Limited